Thirty studies were included. A total of 5,983 patients were included: 663 with CD, 361 with UC, 186 with IBD, 297 with CRC, 697 with IBS and 3,393 healthy controls.
The summary quality scores ranged from 10 to 13. The results of the individual quality assessment were not reported.
All results below refer to a FC threshold of 50 microg/g unless otherwise stated.
IBD compared with no IBD (9 studies, 1,267 patients): the pooled sensitivity and pooled specificity were 89% (95% confidence interval, CI: 86, 91) and 81% (95% CI: 78, 84), respectively. There was strong evidence of heterogeneity (p<0.001). The pooled sensitivity and specificity were both greater at a threshold of 100 microg/g but only 4 studies assessed this threshold.
CD compared with normal controls and IBS (5 studies, 733 patients): the pooled sensitivity and pooled specificity were 95% (95% CI: 92, 97) and 84% (95% CI: 80, 87), respectively. There was some evidence of statistical heterogeneity (p=0.07). Both sensitivity and specificity were greater at a threshold of 100 microg/g but only 2 studies, both restricted to children, assessed this threshold.
UC compared with normal controls and IBS (2 studies, 235 patients): the pooled sensitivity and pooled specificity were 78% (95% CI: 69, 86) and 78% (95% CI: 70, 84), respectively. There was no evidence of statistical heterogeneity (p=0.44).
CRC and adenoma compared with no neoplasia, including inflammation (7 studies, 4,112 patients): the pooled sensitivity and pooled specificity were 36% (95% CI: 34, 39) and 71% (95% CI: 70, 73), respectively. There was strong evidence of heterogeneity (p<0.001).
Adenoma versus no neoplasia, excluding inflammation (3 studies, 547 patients): the pooled sensitivity and pooled specificity were 52% (95% CI: 42, 61) and 77% (95% CI: 73, 81), respectively. There was strong evidence of statistical heterogeneity (p<0.001). Sensitivity was lower and specificity higher in studies that used a threshold of 100 microg/g.
CRC versus no neoplasia, excluding inflammation (4 studies, 2,025 patients): the pooled sensitivity and pooled specificity were 87% (95% CI: 77, 94) and 76% (95% CI: 74, 78), respectively. There was strong evidence of statistical heterogeneity (p=0.007).
The exclusion of lower quality studies increased sensitivity and either had no effect on specificity or decreased specificity. The restriction to large studies had a similar effect. None of the variables investigated in the meta-regression analysis showed a significant association with the diagnostic odds ratio.
The funnel plot suggested a lack of publication bias.